Skip navigation

Anticoagulant decision analysis and guideline


Clinical bottom line

The use of decision analysis methods, based on a systematic review of the literature and incorporating patient values, guidelines can be drawn up that are useful for clinical practice.


R Thomson et al. Decision analysis and guidelines for anticoagulation therapy to prevent stroke in patients with atrial fibrillation. Lancet 2000 355: 956-962.


This study is so complete and detailed that it almost defies précis. Briefly, a number of extensive literature searches were performed to generate information on:

A decision model was used using ahypothetical cohort of patients, with patients able to move from one state to another - as, for instance, if a patient has a major stroke, or bleed, or dies. Information on this was obtained from the systematic reviews. The model was run for 1512 combinations of age, sex, blood pressure and risk factors, and assessed in terms of quality adjusted like years (QALYs) and costs.

A guideline development group that included various medical specialties met to define the scope, and to develop specific questions for review and modelling, and advise on the guideline produced.


Because of the lack of literature information, health utilities were determined by interviews with 57 patients representative of those in clinical practice, if a little younger than average. Low health utility was given to major stroke, and moderate utilities to minor strokes and bleeds (Table 1). Table 1 shows the mean, median and modes on scales of 0 (immediate death) to 1 (normal health).

Table 1: Health utility values associated with anticoagulation

Health state

Utility value




On warfarin with GP




On warfarin with outpatients




Major bleed




Mild stroke




Major stroke




There were four distinct outcomes from the decision model, and these are shown in Table 2, together with the decision made on the basis of each outcome.

Table 2: Model outcomes and clinical decisions

Model outcome

Clinical decision

Treatment produces QALY gains and cost savings

definitely treat

Treatment produces QALY losses and higher costs

definitely do not treat

Treatment produces higher QALYs but with higher costs

treat if cost per QALY is acceptable

Treatment produces QALY losses at lower costs

definitely do not treat

In most cases treatment led to lower costs. In only 12 of the 1512 cases modelled was there a cost per QALY gain. Results for 12 age and sex groups were tabulated using the classification in Table 2, and these are available in the original paper as simple look up figures.

A treatment algorithm was produced for certain clear-cut cases, and with treatment decisions for patients not fulfilling these clear-cut definitions based on the look up tables. The algorithm is shown in Figure 1.

Figure 1: Treatment algorithm

What happens in real patients?

In a consecutive community sample of 207 patients with atrial fibrillation, 116 (56%) had no contraindications to treatment. There were 46 women, 13 aged between 65 and 74 years and 33 aged 75 years or older. There were 70 men, 19 aged between 65 and 74 years and 51 aged 75 years or older.

The proportion of patients above the risk threshold for warfarin treated was almost 100% in each group when the utility of warfarin was 1.0. When it fell to 0.92, the proportion dropped dramatically (Figures 3 and 4).

Figure 3: Proportion of women above the risk threshold for warfarin treatment for different utility values

Figure 4: Proportion of men above the risk threshold for warfarin treatment for different utility values


This is an important piece of work about how guidelines can and should be developed, especially when contrasted with the way in which anticoagulation guidelines have been developed . What this shows is that anticoagulation treatment may be successfully applied to almost all patients, but that that conclusion is exquisitely sensitive to how individual patients see the impact of warfarin therapy on their lives.

It re-emphasises, if that were necessary, that the choices of individuals are extremely important in making decisions about prophylaxis.